How accurate are they for basic sleep metrics like total sleep time and efficiency?
For tracking basic sleep metrics like total sleep time and sleep efficiency, consumer devices are moderately accurate but consistently show a bias toward overestimating how well you slept. A large meta-analysis combining data from 24 studies and 798 participants found that, on average, wrist-worn devices overestimated total sleep time by about 17 minutes and sleep efficiency by about 5%, while underestimating sleep latency (how long it takes to fall asleep) by about 2.5 minutes and overestimating wake after sleep onset by about 13 minutes [1]. This means your device might tell you you slept longer and more efficiently than you actually did.
A separate study testing the Oura Ring, Fitbit Sense 2, and Apple Watch Series 8 against gold-standard polysomnography (PSG) in 35 healthy adults found that all three devices were similar to PSG in estimating total sleep duration, but they showed moderate to substantial disagreement on sleep stages [2]. The Apple Watch, for example, underestimated deep sleep by a striking 43 minutes and overestimated light sleep by 45 minutes [2]. So while the total time in bed might be close, the breakdown of sleep stages can be significantly off.
Can these devices diagnose sleep disorders like sleep apnea?
No, consumer sleep trackers are not reliable for diagnosing sleep disorders like obstructive sleep apnea (OSA), though they may have some screening potential. One study found that a photoplethysmography-based activity device (PAD) significantly overestimated sleep duration and sleep latency compared to PSG, and it could not predict the apnea-hypopnea index (AHI), the key metric for diagnosing sleep apnea [3]. The device's ability to predict AHI was essentially zero (adjusted R² < 0.01), meaning it was useless for gauging respiratory severity [3].
However, there is some evidence that smartwatch data collected over multiple nights at home can be used to predict OSA with moderate accuracy. One study of 59 people found that using the smartwatch's lowest oxygen saturation reading of 85% could predict an AHI ≥5 with 85.7% sensitivity and 70.6% specificity, yielding an area under the curve (AUC) of 0.901, which is considered good [5]. This suggests that while a single night's data from a consumer device can't diagnose apnea, tracking trends in oxygen saturation over several nights might help identify people who need a formal sleep study. For children with syndromic craniosynostosis, ambulatory home sleep studies were found to be reliable for diagnosing OSA in older children, but this used dedicated medical devices, not consumer wearables [6].
Are they useful for tracking trends and predicting behavior?
Yes, despite their limitations for precise measurement, sleep trackers are quite useful for tracking long-term trends and even predicting next-day behavior. A study using a sleep-tracking game app (Pokémon Sleep) with over 2,000 participants found that over 90 days, average total sleep time increased by about 0.8 hours (from 5.5 to 6.3 hours), and improvements in sleep parameters were associated with a decrease in body mass index [8]. This shows that the act of tracking itself can motivate positive change.
More impressively, contactless sleep monitoring technology (using infrared cameras and microphones) was able to predict adverse daytime behaviors in individuals with autism. Over 2,000 nights of data, the system predicted morning adverse behaviors with 71% balanced accuracy, and the effect was strong enough to suggest moderate clinical utility (Cohen's d = 0.45) [4]. This demonstrates that even if a device isn't perfect at measuring sleep stages, the patterns it detects can have real predictive power for health and behavior. Another study showed that data from movement sensors and a sleep tracking device could even detect whether a person with dementia had visitors, with 76.8% accuracy [7].
About These Sources
This answer is built on 8 peer-reviewed studies — published from 2022 to 2026, 6 from 2024 or later, 1 in Q1 journals, collectively cited 56 times — selected as the most relevant from 15 studies that passed quality screening, drawn from 80 papers retrieved from a database of over 500 million.
Sources used in this answer
Performance of consumer wrist-worn sleep tracking devices compared to polysomnography: a meta-analysis
This meta-analysis of 24 studies and 798 participants found that wrist-worn sleep trackers significantly overestimate total sleep time by ~17 minutes and sleep efficiency by ~5%, while underestimating sleep latency and overestimating wake after sleep onset, concluding they are not reliable for clinical use but can track general patterns.
Accuracy of Three Commercial Wearable Devices for Sleep Tracking in Healthy Adults
In 35 healthy adults, the Oura Ring, Fitbit Sense 2, and Apple Watch Series 8 all showed >95% sensitivity for detecting sleep vs. wake, but accuracy for sleep stages ranged from 50-86%, with the Apple Watch underestimating deep sleep by 43 minutes.
0542 Accuracy and Predictive Value of a PAD Sleep Monitoring Device Compared with In-Lab Polysomnography
A photoplethysmography-based activity device (PAD) overestimated sleep duration and latency compared to PSG in 59 nights of data, and could not predict the apnea-hypopnea index (AHI), making it unsuitable for diagnosing sleep-disordered breathing.
0325 Contactless Sleep Monitoring Technology to Predict Adverse Behaviors in Autism
Using contactless infrared cameras and microphones over 2,000 nights in 14 individuals with autism, the system predicted next-day adverse behaviors with 71% balanced accuracy, showing that sleep patterns can predict daytime behavior.
Usefulness of Continuous Sleep Tracking by Smartwatch in Predicting Sleep Apnea
In 59 people, a smartwatch's lowest oxygen saturation reading of 85% predicted an AHI ≥5 with 85.7% sensitivity and 70.6% specificity (AUC 0.901), suggesting multi-night smartwatch data can help screen for sleep apnea.
Accuracy of Detecting Obstructive Sleep Apnea Using Ambulatory Sleep Studies in Patients With Syndromic Craniosynostosis.
In 123 children with syndromic craniosynostosis, ambulatory home sleep studies were found reliable for diagnosing OSA in older children and could guide clinical decision-making, though this used medical-grade devices.
Identifying Social Visitations in a Household of Person Living with Dementia using Movement Sensors and a Sleep Tracking Device
Using movement sensors and a sleep tracking device, a machine learning model could detect whether a person with dementia had visitors with 76.8% accuracy, based on changes in bed and bedroom activity patterns.
Temporal changes in sleep parameters and body mass index after using a sleep-tracking app with gamification.
In 2,063 Japanese adults using a sleep-tracking game app (Pokémon Sleep) for 90 days, average total sleep time increased by 0.8 hours, and improvements in sleep parameters were associated with a decrease in body mass index.